Skip to main content Skip to navigation

Students' Reliance on AI in Higher Education: Identifying Contributing Factors

Project Overview

The document explores the integration of generative AI in higher education, emphasizing how undergraduate students interact with AI tools. It categorizes student reliance on AI into three patterns: appropriate reliance, where helpful recommendations are correctly accepted; overreliance, where flawed suggestions are mistakenly accepted; and underreliance, where beneficial advice is incorrectly dismissed. The findings reveal that students with higher programming self-efficacy, programming literacy, and a strong need for cognition tend to demonstrate more appropriate reliance on AI tools. These insights underscore the critical need to cultivate students' critical thinking abilities and domain knowledge to minimize the risks of overreliance and enhance effective engagement with AI technologies. Overall, the study highlights the potential of generative AI to enhance educational experiences while cautioning against pitfalls associated with its misuse.

Key Applications

AI assistant providing programming problem recommendations

Context: Undergraduate students in computing-related courses

Implementation: Controlled experiment with pre- and post-surveys, and a programming task involving an AI chatbot

Outcomes: Insights into reliance patterns, factors influencing reliance, and potential interventions to promote appropriate reliance on AI

Challenges: Overreliance leading to acceptance of incorrect AI suggestions and underreliance causing rejection of helpful recommendations

Implementation Barriers

Cognitive Bias

Automation bias leads students to favor AI suggestions over their own judgment, potentially undermining learning outcomes. Lack of transparency in AI decision-making creates trust problems, making students vulnerable to overreliance.

Proposed Solutions: Educational interventions that promote critical evaluation of AI outputs and build domain expertise. Implementing cognitive forcing functions and reflective prompts to enhance critical engagement with AI.

Project Team

Griffin Pitts

Researcher

Neha Rani

Researcher

Weedguet Mildort

Researcher

Eva-Marie Cook

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Griffin Pitts, Neha Rani, Weedguet Mildort, Eva-Marie Cook

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies